Zanarotti@mit-multics.arpa (02/18/86)
Sendmail appears to be somewhat broken in 4.3BSD. Ever since ATHENA went to 4.3, we began receiving returned mail like the ones you got. The "replacement out of range" is a result of an incompatible change to the ruleset evaluator. In 4.2, sendmail did not complain if the replacement specified a string argument that did not exist. For instance, if the ruleset matched on three strings, and you gave it $4 in the replacement, sendmail would not complain. As a result, there are buggy configurations out there, that sendmail is now complaining about. The "endmailer: no children" message is an intermittent error that has been plaguing us for about a month. Looking at the code, it looks like sendmail is waiting for the delivery mailer, but the PID it's waiting for doesn't exist. This could be a kernel problem. Has anybody tracked this one down?